Databases of the Alberta Kidney Disease Network (http://www.akdn.info; Alberta, Canada) were used to identify a cohort aged 66 and older with diabetes who were incident metformin users from November 1, 2002 until March 31, 2008. The cohort was stratified on frequency of creatinine measurements prior to initiating metformin: none; a pre measurement only (6 months prior to metformin initiation); a post measurement only (within 1 year after); and both pre and post measurements. Patients were categorized by stage of CKD: 1 & 2 ( ≥ 60 mL/min/1.73m2; reference); 3a (45 - 59 mL/min/1.73m2); 3b (30 -44 mL/min/1.73m2); 4 & 5 (< 30 mL/min/1.73m2); and no measurements. A Cox proportional hazards model was used to determine the association between kidney function and mortality, controlling for demographics, Charlson morbidity scores, hypertension, and visits to general practitioners.